Partisan consensus and divisions on content moderation of misinformation

Cameron Martel; Adam Berinsky; Paul Resnick; Amy X. Zhang; David Rand

In brief

Despite a partisan gap, majorities of both Democrats (93%) and Republicans (65%) support tech platforms reducing the spread of misinformation. Partisan divisions emerge on specific interventions: 85% of Democrats but only 46% of Republicans support banning frequent misinformation sharers. A conjoint experiment shows both parties prefer expert juries for content moderation decisions, though Republicans rate experts as less legitimate. Critically, large, informed layperson panels can achieve comparable legitimacy to expert panels when juries are large, well-qualified, and allowed deliberation.

Contributions

First nationally representative survey (N=2,909) directly measuring American public opinion on content moderation of misinformation across the political spectrum.
Empirical evidence that core agreement exists on whether platforms should moderate (80% overall support), even if partisan gaps exist on how.
Conjoint experiment measuring perceived legitimacy of different jury compositions (expert vs. layperson, size, qualifications, deliberation) for deciding whether content is harmful.
Demonstrates that procedural legitimacy—not just efficacy—is a constraint on platform moderation policy; partisan distrust of expertise is real but does not preclude technical solutions (e.g., scaled deliberative panels).

Method

Nationally representative poll on YouGov (N=2,909; July-August 2023) with three components:

Perception of harmfully misleading content: 7-point Likert scale measuring (a) whether harmfully misleading content is a problem for society, and (b) whether it is dangerous to the public. Partisan breakdown: Democrats (M=5.9) vs. Republicans (M=4.6) on danger; (M=5.0) vs. (M=4.3) on problem severity.
Support for platform intervention: Binary question: "Should social media companies try to reduce the spread of harmful misinformation?" Secondary question for affirmative responses: "Should platforms ban users who frequently share harmfully misleading content?"
Legitimacy conjoint experiment: Respondents evaluated multiple hypothetical moderation juries that varied on composition (experts vs. laypeople), size (smaller vs. larger), qualifications (minimum knowledge requirements), and process (whether discussion was allowed). Outcomes measured perceived legitimacy of each jury configuration.

Results

Problem perception: Americans on average rate harmfully misleading content as a "notable problem" (M=4.9) and "somewhat dangerous" (M=5.2), comparable to depictions of violence (M=5.1) and hate speech (M=5.4). Partisan gap is substantial but both groups acknowledge the problem.
Support for reduction: 80% of all respondents support platforms reducing misinformation (93% Democrats, 78% Independents, 65% Republicans).
Support for banning frequent sharers: 67% overall support banning frequent misinformation sharers (85% Democrats, 61% Independents, 46% Republicans). This is the largest partisan divide in the survey.
Jury legitimacy (conjoint experiment):
Experts perceived as most legitimate across all political groups.
Republicans rate expert juries as less legitimate than Democrats but still more legitimate than baseline layperson juries.
Size, qualifications, and deliberation each increase perceived legitimacy of layperson juries.
Maximally legitimate layperson juries (large, well-qualified, with discussion enabled) are comparably legitimate to expert panels across the political spectrum.

Notes

Strengths: Large nationally representative sample; addresses a critical policy gap (what does the public actually want?); conjoint experiment isolates specific procedural components of legitimacy; directly challenges polarization narratives by documenting broad consensus on the problem and general goal (reduce misinformation) even where partisan divide emerges on specific mechanisms.

Limitations: Survey conducted July-August 2023; landscape of platform moderation practices shifted substantially afterward (platforms rolled back policies in 2024 ahead of election). Conjoint experiment measures hypothetical jury legitimacy, not revealed preferences in actual moderation systems. No mechanism analysis: what drives layperson jury legitimacy improvements (greater diversity? perceived deliberation? larger number = perceived representativeness?) remains unclear. No causal analysis of how these preferences affect actual platform behavior or policy adoption.

Policy implications: Results suggest public demand exists for continued platform moderation (contra rollback narratives), and that concerns about partisan bias in moderation can be addressed through procedural legitimacy (transparent jury composition, deliberation, size, qualifications) rather than blanket prohibition or abandonment of fact-checking. Republican support for expert moderation (while lower than Democratic support) suggests policy options based on expert-layperson hybrid approaches may find broader legitimacy.

How it connects

Builds on Emphasizing publishers does not effectively reduce susceptibility to misinformation on social mediaDias et al. · The Harvard Kennedy School Misinformation Review · 2020 Related Partisan Polarization Is the Primary Psychological Motivation behind Political Fake News Sharing on TwitterOsmundsen et al. · American Political Science Review · 2021 Builds on Birdwatch: Crowd Wisdom and Bridging Algorithms can Inform Understanding and Reduce the Spread of MisinformationWojcik et al. · arXiv · 2022 Motivated by Political motives help rather than hinder crowdsourced fact-checkingMartel et al. · Working paper · 2025 Related The platform governance triangle: conceptualising the informal regulation of online contentGorwa · Internet Policy Review · 2019